Command A
Language Available Full comparison ↗ cohere/command-a · by Cohere
· decoder-only-transformer
Pricing — 1 offering(s)
Input tokens
- $2.50 / 1M tokens (input) Current 2025-03-13 → present
Output tokens
- $10.00 / 1M tokens (output) Current 2025-03-13 → present
Showing the active price and any recorded history. Full pricing history is available via the paid API — see API docs.
See how Command A fits into a cost-aware routing setup
See how →Capability profile
How Command A rates across core capability dimensions, with the task-level evidence behind each rating.
Competent at enterprise workloads; not designed for frontier reasoning tasks.
Cohere Command A blog (cohere.com/blog/command-a): MBPPPlus, SQL (BirdBench/Spider Dev/Test), and RepoQA coding benchmarks published; claims 'at the level of GPT-4o on MBPPPlus'. Coding is an explicit primary evaluation axis.
Native multi-step tool use; one of Cohere's core product differentiators.
Strong grounding and instruction adherence; purpose-built for agentic enterprise use.
256K context — 2× Command R+ and best-in-class for Cohere models.
Coding benchmarks
View the full coding benchmark leaderboard →| Benchmark | Score | Harness | Source |
|---|---|---|---|
| swe-bench-verified ↗ | 7.8% 2026-09 | mini-swe-agent | 📊 leaderboard ↗ |
| livecodebench ↗ | 35.1% 2026-09 | — | 📊 leaderboard ↗ |
Operator guidance
Route here for enterprise RAG or agentic workflows where Cohere's grounding capabilities, data-residency options, and enterprise support are valued. At the same price as Command R+, Command A offers double the context and improved tool-use. GPT-4o is a competitor at the same $2.50 / $10 price.
Use cases
- Enterprise RAG pipelines requiring long-context and reliable tool use
- Multi-step agentic workflows with structured grounding
- Document analysis over very long corpora (256K window)
Limitations
- Weaker general reasoning than GPT-4o at the same price
- Capability ratings are qualitative, not a single cited benchmark run